Scalably distributed SystemC simulation for embedded applications

Abstract

SystemC becomes popular as an efficient system-level modelling language and simulation platform. However, the sole-thread simulation kernel obstacles its performance progress from the potential of modern multi-core machines. This is further aggravated by modern embedded applications that are getting more complex. In this paper, we propose a technique which supports the geographical distribution of an arbitrary number of SystemC simulations, without modifying the SystemC simulation kernel. This technique is suited to distribute functional and approximated-timed TLM simulation. We integrate this technique into a complete MPSoC design space exploration framework and the improvement gained is promising.